Igris — Voice AI Agent
A ground-up conversational AI agent — real-time voice in and out, dual-LLM reasoning, and the ability to control the machine it runs on. Closer to a personal GPT than a chatbot widget.
Brief
A self-directed project to build a voice-driven AI agent — not a wrapper around a chat widget, but a real system with its own server, live voice I/O, and the ability to act on the machine it runs on.
Process
An Express + WebSocket server pairs with a React client for real-time streaming, backed by both the OpenAI API and Google's Gemini API for reasoning, with ElevenLabs handling natural speech synthesis — similar in spirit to ChatGPT's voice mode, but built entirely from scratch with its own machine-control layer.
Client
Personal Project
Role
Built by Thinkova — server architecture, dual-LLM integration, and voice pipeline.
Tech Stack
Outcome
A working voice agent that can hold a spoken conversation and control the local machine on request — published on GitHub.